可分的3d卷积比张量流中的3d卷积引起更多的内存？

Question

我想将3D卷积分为（2 + 1）D，这意味着先进行空间卷积，然后再进行时间卷积，即kernel =（D，H，W）-> kernel（1，H，W ）和kernel（D，1,1）。从理论上讲，它具有较少的参数，并且应该节省内存。例如。具有3通道输入和64输出通道的内核（7x7x7），#params = 7x7x7x3x64 = 65,856;可分离的3d：1x7x7x3x64 + 7x1x1x64x64 = 38,080。

但是，在我的测试中，unit_sep3D比unit3D引起的内存更多。为什么会这样？

def unit3D(inputs, output_channels,
           kernel_shape=(1, 1, 1),
           strides=(1, 1, 1),
           activation_fn=tf.nn.relu,
           use_batch_norm=True,
           use_bias=False,
           padding='same',
           is_training=True,
           name=None):
  """Basic unit containing Conv3D + BatchNorm + non-linearity."""
  with tf.variable_scope(name, 'unit3D', [inputs]):
    net = tf.layers.conv3d(inputs, filters=output_channels,
                            kernel_size=kernel_shape,
                            strides=strides,
                            padding=padding,
                            use_bias=use_bias)
    if use_batch_norm:
        net = tf.contrib.layers.batch_norm(net, is_training=is_training)
    if activation_fn is not None:
        net = activation_fn(net)
  return net

def sep3D(inputs, output_channels,
          kernel_shape=(1, 1, 1),
          strides=(1, 1, 1),
          activation_fn=tf.nn.relu,
          use_batch_norm=True,
          use_bias=False,
          padding='same',
          is_training=True,
          name=None):
  """Basic Sep-Conv3D layer with BatchNorm + non-linearity.
  A (k_t, k, k) kernel is replaced by a (1, k, k) kernel and a (k_t, 1, 1) kernel
  """
  k_t, k_h, k_w = kernel_shape
  if type(strides) == int:
    s_t, s_h, s_w = strides, strides, strides
  else:
    s_t, s_h, s_w = strides
  spatial_kernel = (1, k_h, k_w)
  spatial_stride = (1, s_h, s_w)
  temporal_kernel = (k_t, 1, 1)
  temporal_stride = (s_t, 1, 1)
  with tf.variable_scope(name, 'sep3D', [inputs]):
    spatial_net = tf.layers.conv3d(inputs, filters=output_channels,
                                  kernel_size=spatial_kernel,
                                  strides=spatial_stride,
                                  padding=padding,
                                  use_bias=use_bias)
    if use_batch_norm:
        spatial_net = tf.contrib.layers.batch_norm(spatial_net, is_training=is_training)
    if activation_fn is not None:
        spatial_net = activation_fn(spatial_net)
    temporal_net = tf.layers.conv3d(spatial_net, filters=output_channels,
                                  kernel_size=temporal_kernel,
                                  strides=temporal_stride,
                                  padding=padding,
                                  use_bias=use_bias)
    if use_batch_norm:
        temporal_net = tf.contrib.layers.batch_norm(temporal_net, is_training=is_training)
    if activation_fn is not None:
        net = activation_fn(temporal_net)
  return net